Sign in with
Sign up | Sign in
Your question

Ramdisk

Last response: in Storage
Share
December 14, 2011 10:25:48 AM

With current ram prices having hit an all-time low why isn't there any more mention of using ramdisks?
32gb of high performance memory cost as little as 250€, making say a 24gb ramdisk easy and cheap to realize.

Benchmarking a ramdrive appears to be more an exercise in testing benchmark programs and filesystems that are unsuited for a throughput flat out off the scale (many of the benchmarks have to be run in multiple parallel instances before the individual measurements significantly degrade), making even the best ssd's look s-l-o-o-o-o-o-w. And that's with freeware drivers, most having not been updated in ages.

Even volatility is hardly a serious issue with implementations of loading and saving disk images at startup and shutdown. (especially when loading from an ssd).
So ... why isn't there anything? No marketability for a pure software implementation? Using bog standard ram ought to annihilate things like caching ssd's in performance.

More about : ramdisk

a c 353 G Storage
December 14, 2011 11:23:20 AM

Ram drive, free for upto 4 gigs, $15 for > 4 gigs . I have an 8 gig ram drive set up that I can turn on/off as needed. Have 16 gigs of ram.
http://memory.dataram.com/products-and-services/softwar...
Yes, AS SSD will show about a 10x performance boost over an SSD. SSD = Model T ford, Ram Disk = 2011 sport car.

1) Most systems max out at 16 gigs ram an can only move up to the 32 gig limit if 8 gig modules, expensive compare compared to 4 gig modules.
2) Not sure I'd want to have the image stored on an MLC SSD due to the High write cycle. In my case that would be 8 gig write everytime I cycled the system. If you have the periodic Backup enabled the write cycles could go thru the roof. Would recommend a small SLC drive for storing the image.
3) If image is on a HDD, then Both Boot Time and shut down time are increased. Not really an issue in 24/7 operation, or if you turn on in AM and Off at night. Could be an issue if you periodicly cycle during the day.
4) Size could be a limiting factor for editing/encoding very large Video files such as bluray where a single file can be upto 40 gigs (typically 13-> 25 gigs). Would be ok for DVDs where file size is 1 gig per vob file (6 to 7 gigs for entire movie). For gamers, not sure but probably could install maps here whit almost instant load time.

My first Ram disk was with a 386SX system MAX MEMORY = 1 MEG, kind of limited the size of ramdrive.. Back then you could not save the image, so lost everything if powered off (ie lock up and used the 3 finger salute to reboot). My first store bought 386-sx was a rare exception. Could reboot and ramdrive would still be intack, Untill I need to update the Rom Bios (replace the socketed two chips. Would not except new chips and sent it back for repair. After repair, lost ramdrive with a reboot - LOL.
December 15, 2011 6:33:50 AM

I'm currently using a 1GB Ramdisk for certain tasks such as temp files, internet cache, page filing. The question I have is can you use a Ramdisk for HDD caching purposes much like Intel is doing with SRT and SSDs? If so, can you provide a link that instructs a person on how to set this up.
Related resources
December 15, 2011 8:50:04 PM

a4mula said:
The question I have is can you use a Ramdisk for HDD caching purposes much like Intel is doing with SRT and SSDs? If so, can you provide a link that instructs a person on how to set this up.


It's definitely doable, but the only implementation I know of is Superspeed's. It's expensive though and no idea how well, if well at all, it performs. Just google for superspeed supercache and you should find it, they have a fully functional trial version available. Maybe you can try it out and tell us how it works?

I'm looking to buy an -E setup since my current one has just about reached the end of its useful lifespan and since ram amount and performance is really no issue at all there, I guess my post was a bit of a rant about no ramdisk drivers or filesystems that are remotely suitable for a current system like that.
Not that I know of any disk benchmark programs that aren't left gasping in the dust even by older implementations and far less capable systems. Are there any?
There's almost no data available that isn't years out of date and wildly inaccurate to boot.
It'd almost be worth it to make a little DIY roundup.
December 19, 2011 7:34:13 PM

more then 4GB memory base one use 1~2GB ramdisk.
Make sure temp and history cache is on ramdisk.
your pc is fly~ fly~
December 23, 2011 4:40:29 PM

SuperSpeed SuperCache works great! I've only been using it a couple of days but I'm impressed. I already had a 4GB RamDisk dedicated to Paging as my only pagefile for all disks and that in itself was pretty impressive performance wise, that completely eliminated all micro-stutter from my HD6870 crossFire set up.

A few days ago I added a 8GB RamCache (SuperSpeed SuperCache). The way it works is it dedicates the amount of ram you choose to a specific disk for a cache. I dedicated it to my sata II SSD raid 0 array that runs my OS and programs. The increase in speed and performance is very noticeable.

I have 24GBs of ram 4 for my RAM pagefile, 8 for my RAM Cache and 12 left over for the OS. If you have enough ram try it. It definitely makes a bigger difference than I thought possible.

Here's a screenshot of AS SSD. Just click on the images if you want to see them larger.


And one of CrystalDiskMark


:sol: 
December 23, 2011 5:15:42 PM

I've never heard of a RAMDisk effecting microstutter issues. I can see how it'd speed up texture caching for files coming from the storage subsystem, but I'm not really sure how it would change the Alternate Frame Rendering micro-stutter associated with crossfire/sli setups. This is more of a synchronization issue where one card is rendering a screen faster than the second creating a lag in the output.

Wow, those are some smokingly fast reads and writes. I'm suprised you didn't choke the Marvell controller. Are you running those off a discrete RAID controller?

Either way, thanks for posting your results. I'll definitely be giving that program a look.

Now I'm curious if you could use that as a pre-buffer for SRT. A tiered approach at caching. I'm weary of my current SSDs, just the constant buzzing in the back of my head about write thrashing, space issues, and the pain that setting SSDs up right the first time really has me leaning towards the install and forget nature of it.

Thanks again, really. Very insightful. Great rig also.
December 23, 2011 6:09:11 PM

a4mula said:
I've never heard of a RAMDisk effecting microstutter issues. I can see how it'd speed up texture caching for files coming from the storage subsystem, but I'm not really sure how it would change the Alternate Frame Rendering micro-stutter associated with crossfire/sli setups. This is more of a synchronization issue where one card is rendering a screen faster than the second creating a lag in the output.

Wow, those are some smokingly fast reads and writes. I'm suprised you didn't choke the Marvell controller. Are you running those off a discrete RAID controller?


I hear what your saying about micro-stutter which is why I took a year to try that solution. I really don't know exactly why it worked only that some others claimed it did and eventually as a last resort i tried it. All I can say for sure is it completely solved the problem.

No discrete RAID controller, Just the on-board Intel controllers. I use the Marvell controllers for my optical drives only (not to fond of my 1st gen Marvell controllers).

The app (SuperSpeed SuperCache) seems to be very well written. With read every thing you need to really speed things up gets cached and with write there is an option for "lazy Write" which can delay a write (keep it in ram) for anywhere from 1-15 seconds (you choose) I have it set to 5 sec. This really speeds up writes but could be especially dangerous for raid 0 without a good battery back-up.

So far it seems to be a really good app and THX for the PC kudos. Yours isn't to shabby either :sol: 
December 23, 2011 6:51:14 PM

Idonno said:

No discrete RAID controller, Just the on-board Intel controllers. I use the Marvell controllers for my optical drives only (not to fond of my 1st gen Marvell controllers).


I feel you on the Marvell controller. So.. that means you're running ICH10R 3Gb/s? I was under the impression it capped at 660MB/s. Interesting results. Not sure how to look at it, transfer rates from the QPI are exponentially higher, so this could be factoring not only transfer rates of the RamDisk to SSD (ICH10R), but also CPU to RamDisk (QPI) which could skew the results.

Hmm, I dunno. Will have to consider what's going on here more. At some point the data has to pass through ICH10R which is going to be completely saturated creating a bottleneck. I'd really love to see this setup on an Areca 1880 or at a minimum a decent pci-e SATA 6Gb/s card.



a c 353 G Storage
December 23, 2011 8:24:30 PM

Ramdrive performance is higher, ie Seq reads in the area of 5000. Other values are also higher. Currently on vacation (LA) so can not post a screen shot. With Ramdrive the Sata controller is irrelevant as Performance is a function of memory
December 23, 2011 10:48:42 PM

RetiredChief said:
Ramdrive performance is higher, ie Seq reads in the area of 5000. Other values are also higher. Currently on vacation (LA) so can not post a screen shot. With Ramdrive the Sata controller is irrelevant as Performance is a function of memory


Yea, but your comparing apples to oranges.

If I was just posting screen shots of a RamCache in action I think the speeds would be similar since they both use my ram.

This is my 4GB RamDisk


The screen shots I posted earlier were of my 240GB C drive assisted by a 8GB RamCache which is very different and much more practical than a RamDisk in many ways.

A RamDisk for the purpose of a C drive would not be practical anyway, unless you only wanted a very small drive and you never planned on shutting down your PC. It would take far to long and the first time you lost power or had to do a hard shutdown your OS would be toast. :D 
a c 353 G Storage
December 24, 2011 12:32:17 AM

@ Idonno. Got you.
However Benchmark programs are not very representative of performance for large "cached" HDD as real life performance is highly dependent on algorthium that does the cache and also how repetitive file reads are.
December 24, 2011 7:34:24 AM

RetiredChief said:
@ Idonno. Got you.
However Benchmark programs are not very representative of performance for large "cached" HDD as real life performance is highly dependent on algorthium that does the cache and also how repetitive file reads are.


That may be true, but unlike my opinion ("It definitely makes a bigger difference than I thought possible"), benchmarks do offer real evidence.

It's not a "large "cached" HDD", It is a fairly small hard drive (240GB) assisted by a very large cache (8GB RamCache)

While it's true an 8GB cache would not offer anywhere near as much benefit on large sustained Reads/Writes of 8GBs and greater, the speed of your OS and programs are almost entirely dependent on much smaller Reads/Writes.
My over-all access time seems to have been improved as well.

I personally give this program an A+. At $79 (I think) it's not cheap, however in my opinion It offers me a better real life performance increase than adding another 1 or 2 SSD's to my existing raid 0 array so the price is worth it.

Will it help you with a large file transfer, say maybe a 40GB blu-Ray? Not so much. But even if you had 5 Sata III SSD's in raid 0, unless you had somewhere just as fast to transfer it to or from that wouldn't help you either.

So if I do decide to purchase it, I think it will be money well spent. Right now My OS feels incredibly snappy! I like it, but the only way for you to know is to try it. It's free for 14 days.

a4mula said:
Hmm, I dunno. Will have to consider what's going on here more. At some point the data has to pass through ICH10R which is going to be completely saturated creating a bottleneck. I'd really love to see this setup on an Areca 1880 or at a minimum a decent pci-e SATA 6Gb/s card.


That's true but as long as the read/write activity isn't sustained over too long a period the RamCache should be able to buffer this. A real limiting factor here is the size you choose for your RamCache and your available ram, although a fast raid card or a fast RevoDrive would certainly improve the RamCache's ability to buffer the Read/Writes. :sol: 
a c 353 G Storage
December 24, 2011 6:44:56 PM

^ "got you" was a poor choice of words - should have said Understand you.
Always nice to have a 2nd opinion. And with a free 14 day trial, it's always worth a shot.
December 24, 2011 6:57:03 PM

RetiredChief said:
^ "got you" was a poor choice of words - should have said Understand you.
Always nice to have a 2nd opinion. And with a free 14 day trial, it's always worth a shot.


That's OK even though It could be taken a few different ways I knew what you meant.

And yes, you should give it a try. It's very easy to implement and even if you decide you don't want it, it should be kind of fun anyway. :D 
a c 415 G Storage
December 25, 2011 1:44:38 AM

IMHO using main memory as a RAM drive is largely a waste of time unless you have a very specific need and have confirmed that the end performance really is better. And I'm not talking about synthetic benchmarks, I'm talking about timing some activity (like playing a game or editing a video) to see if a RAM cache actually makes a measurable difference.

A problem with RAM drives is that they loose their data when powered off - this means that you need to load them up with data, typically from a hard drive, when the system is powered up anyway. When you combine that with the fact that Windows does a pretty good job of caching disk activity in unused RAM anyway, a RAM drive doesn't really buy you very much.

Take temp files, for example. When a program writes to a temp file to a hard drive, Windows caches the writes - so the program sees its output operations complete instantly while Windows does "lazy writes" in the background to dump the data to the drive. Then when the program comes back to read the data again, Windows supplies it from the cache and again the I/Os complete instantly.

The only instance I can see in which a RAM drive would be beneficial is if you have specific critical data you want to keep in RAM even though you're doing heavy reading and writing of other files at the same time to the extent that your critical data gets kicked out of the cache. In effect the RAM cache gives you the choice of forcing some data to remain in RAM at the expense of other data, something that you don't have control over otherwise.

But that's not really a scenario that applies to most people.

Having said all that, the original premise - buy lots of RAM for caching - still applies. It's cheap, and it really does make a difference when you're going a lot of repetitive file accesses, even if you just let Windows use it for caching.
December 25, 2011 7:01:04 AM

@sminlal You seem to be getting RamDrive & Ram Cache mixed up. While they do have very real similarity's they are not the same and they serve very different purposes.

Windows only does a pretty good job of caching disk activity in unused RAM when you have less than about 12 GB's of total system ram. At this time windows is still not written to take advantage of the larger amounts of ram that the newer motherboards are capable of.

A very good example is if you have 24-64GB's of ram and set your Paging file to "let windows Choose" instead of using ram or even offering that as a alternative windows will set a fairly large Paging file on all of your hard drives.

In fact whether you have 12 or 24 GB's of ram windows will use approximately the same amount. This isn't because at 12GB's windows is already using the optimal amount. This is because windows has not yet been written to take advantage of the larger amounts of ram recently available to the average user.

"synthetic benchmarks" while I do agree that they don't ALWAYS properly represent ALL real life usage. They still are one of the best tools for checking hardware performance that we have. For instance if you check the users configuration in any of the top PC benchmark programs that store these statistics you will not find a single crappy PC in any of the top slots, similarly all you will find at the bottom are crappy PC's and PC's with hardware issues.

Your statement:
sminlal said:
Having said all that, the original premise - buy lots of RAM for caching - still applies. It's cheap, and it really does make a difference when you're going a lot of repetitive file accesses, even if you just let Windows use it for caching.

Is true if by lots of ram you mean 8 or 12GB. however by today's standards with a few LGA 2011 motherboards capable of 128GB's and still more capable of 64GB's of ram, 12GB's is at least starting to seem rather small. And I assure you without a program like SuperSpeed SuperCache to assist windows won't use all that extra ram for caching or anything else.

Maybe win 8, we'll see, but certainly not 7-64 or any previous windows. :sol: 

P.S. I just did a little test: I created a 4.36 GB ISO image file from DVD files. Total Time: 36 sec. :D 
a c 415 G Storage
December 25, 2011 5:30:57 PM

Idonno said:
@sminlal You seem to be getting RamDrive & Ram Cache mixed up. While they do have very real similarity's they are not the same and they serve very different purposes.
No, I'm not mixing them up, what I'm saying is that since Windows' RAM cache keeps a copy of the data you're using in RAM anyway, using a RAMDrive to manually load stuff into RAM doesn't really give you any particular advantage except in certain specific scenarios.

Idonno said:
A very good example is if you have 24-64GB's of ram and set your Paging file to "let windows Choose" instead of using ram or even offering that as a alternative windows will set a fairly large Paging file on all of your hard drives. In fact whether you have 12 or 24 GB's of ram windows will use approximately the same amount.
I'm not sure why you're discussing the page file size here - it really doesn't have anything to do with caching file accesses in RAM. In fact almost everyone with 12GB or more of RAM (and even less) would probably be better off simply disabling their page file altogether. The "classic" page file sizing rule of 1.5X your RAM size makes no sense whatsoever when you get up into these kinds of large RAM sizes, and I'm sure Microsoft was smart enough to put some sort of cap on the default maximum page file size. I've run with no page file for a few years now and my system works just great.

When you have, say, 24GB of RAM there are only three reasons I'm aware of why Windows wouldn't be using it for file caching:

1) if it's being used by some other program,

2) if the program doing the file accessing is bypassing the cache (for example, backup or integrity checking programs often do this), or

3) if you haven't actually accessed 24GB's worth of data in your files yet. That's a lot of space, and it takes quite a bit of file activity to actually access that much data. It's only when you start copying around huge files like movies that it starts to take advantage of that much RAM cache. And even then, you only get the benefit the second or subsequent times you read the file (although write caching is an immediate benefit).
December 25, 2011 7:39:27 PM

sminlal said:
No, I'm not mixing them up,

In that case, I apologize for telling you what you seem to think instead of asking you what you thought/meant. That was rude and that very thing has pissed me off in the past as well. I'm sorry!
But for the sake of fairness, if you re-read your post I think you might see how at least I might have jumped to that conclusion.

sminlal said:
I'm not sure why you're discussing the page file size here - it really doesn't have anything to do with caching file accesses in RAM.

Your right it doesn't, however it is a perfect example of windows not making use of excessive ram

sminlal said:
I'm sure Microsoft was smart enough to put some sort of cap on the default maximum page file size.

At 24GB's of ram I haven't seen it yet. Maybe someone with more ram can chime in here.

sminlal said:
I've run with no page file for a few years now and my system works just great.

That's great. I tried it for a while and for the most part everything was ok. However it should be noted that there are some applications that require a page file and after I created a 4GB RamDisk and dedicated it solely for my pagefile things ran noticeably smoother. (yet another reason to bring up pagefile)

sminlal said:
When you have, say, 24GB of RAM there are only three reasons I'm aware of why Windows wouldn't be using it for file caching

I'm not going to even try to argue the finer points as to why windows doesn't use excessive ram for ANYTHING, only that from personal experience it doesn't! Even with 53 instances of WinRar running at the same time with 24GB's of ram the most it used was a little over 8GB's. If you want to use excessive ram (over 12GB's of system ram) you need to use an app that is specifically designed to take advantage of large amounts of ram.

I am not speaking of the way I think things should be. I am speaking from experience about the way I have found things to be.

I have used 24Gb's of ram for over 1yr now. RamDisk for 1 month and SuperSpeed SuperCache (RamCache) for just under a week. I see value in both especially RamCache. I came to these conclusions through experience with these app's and the noticeable difference in performance they provided. NOT because some benchmark told me my system was now faster, all the benchmark did (for me) was point out what was already obvious.

Lastly and most important MERRY CHRISTMAS to all :D 
a c 415 G Storage
December 26, 2011 4:53:00 AM

Idonno said:
Even with 53 instances of WinRar running at the same time with 24GB's of ram the most it used was a little over 8GB's.
That's not a limitation of Windows, it's how WinRar works. When you're copying files around you really can't get any faster than using double buffering for the input and output files - that way you always have an I/O outstanding to each drive and the CPU won't be a bottleneck. It doesn't take very much RAM to do that, only a few megabytes per copy thread.

Windows will very happily allocate tons of RAM to a program if it asks for it. When I decided to run without a pagefile I wrote a program which sat around allocating memory until the system ran out of it just to understand what would happen. Windows didn't have any qualms about granting as much memory to my program as it asked for, promptly ran out of RAM, and displayed a dialogue box which offered to cancel the offender. Once the wayward program was cancelled, the system carried on as if nothing had happened.

The memory manager in 64-bit Windows 7 is exactly the same one that's used in Windows Server 2008 R2. The only difference is the fact that some artificial caps on memory size are removed from the server version, and the default settings favour background services over foreground GUI programs. You can be sure that Microsoft would have some very unhappy corporate customers if Windows didn't let them fully utilize the huge amounts of very expensive ECC server memory they cram into their servers.

Idonno said:
it should be noted that there are some applications that require a page file and after I created a 4GB RamDisk and dedicated it solely for my pagefile things ran noticeably smoother.
Yes, I had that experience with Photoshop V6.0. It's one of the reasons I upgraded to CS4, which works very happily without a pagefile. Those applications are, fortunately, getting fewer and farther between.

I agree that if you've got one of those programs then yes, you do need to create a pagefile. But it doesn't have to be a huge file, you can create tiny little 10MB pagefile and the system and the wayward program will both work just fine. In fact with the kind of RAM we're talking about, the system isn't going to be putting anything significant into the pagefile anyway, so you're pretty unlikely to see any noticeable performance degradation if you just let the system create it on the OS drive, even it it's a hard disk. The one thing to watch out for is if the wayward program is using the pagefile location to decide where to put its work files - in that case the location of the pagefile does matter, but only indirectly.
December 26, 2011 9:05:15 AM

WinRar was just one of many possible examples. And I really don't see the point in arguing with you any more. You say things like:"Windows will very happily allocate tons of RAM to a program if it asks for it" OK so whats your point?

Because I said "If you want to use excessive ram (over 12GB's of system ram) you need to use an app that is specifically designed to take advantage of large amounts of ram." Why? Because windows wont allocate large amounts of ram for ANYTHING unless it is asked to by "an app that is specifically designed to take advantage of large amounts of ram." Wow, didn't I say that before? It sounds kind of like what you said only with slightly different words!

So go and try RamDisk for a pagefile or RamCache for your OS drive (you might want to get some more ram first) then get back to me. Your credibility then will be greatly enhanced but right now you sound kind of like one of history's naysayers. Just so you know, the world isn't flat, more than 15mph on a train won't result in a heart attack, man can fly even though god didn't give him wings and yes man can and has walked on the moon.

The point is, myself and quite a few others have tried RamDisk and RamCache and have realized the benefits first hand. You on the other-hand seem so set in your belief that they provide very little benefit, you will not even listen to people that have or god forbid try it yourself.

That leaves you with very little credibility on this issue and quite frankly I am unwilling to waste anymore of my time arguing with you! Try it or don't. I don't care. :kaola: 
December 26, 2011 3:27:47 PM

Thank you again for posting your results and experience with the product Idonno. I don't claim to be technical enough to intelligently contribute to the debate but I can honestly say that at $79 dollars and with a 2 week free trial it will be something that will at least be examined in my new build.
December 26, 2011 4:11:40 PM

a4mula said:
Thank you again for posting your results and experience with the product Idonno. I don't claim to be technical enough to intelligently contribute to the debate but I can honestly say that at $79 dollars and with a 2 week free trial it will be something that will at least be examined in my new build.


Thanks a4mula, I'm glad you found the information at least provocative. I was starting to think I made a mistake by even posting here but I guess as long as a few found my info interesting I didn't. :pt1cable: 
a c 415 G Storage
December 26, 2011 5:41:51 PM

Idonno said:
Because I said "If you want to use excessive ram (over 12GB's of system ram) you need to use an app that is specifically designed to take advantage of large amounts of ram." Why? Because windows wont allocate large amounts of ram for ANYTHING unless it is asked to by "an app that is specifically designed to take advantage of large amounts of ram." Wow, didn't I say that before? It sounds kind of like what you said only with slightly different words!
Actually, what started me out on this was your first reply to my first post, in which you said:

Quote:
At this time windows is still not written to take advantage of the larger amounts of ram that the newer motherboards are capable of.

My point is that Windows has no such limitation. If you've got the memory, Windows will use as much of it as it needs for file caching and it will let programs use as much as they need. WinRAR simply doesn't need a lot of memory, so it's a mistake to use it as an example of supposed problems with Windows memory management. Most other programs are the same. It's only extremely memory-intensive programs such as video editing, database serving and the like which can actually do something useful with that kind of memory.

Saying that Windows has a memory management problem because programs don't allocate a lot of memory is like blaming GM for low ridership when a transit company uses their 40-seat buses on a route that only gets a few customers.
December 26, 2011 7:26:30 PM

Nope I'm not going to do it, I could......... no, I'm just not going to! :pt1cable: 
December 27, 2011 2:22:35 AM

If you want a RAM disk then use imdisk, it's free and opensource. Works with 32 and 64 bit Windows.



I have a rather good implementation of a RAM disk at home.

16GB of DDR3-1600 memory, 4 x 4GB sticks.

I first used imdisk to make a 64kb ram disk with a file image (writes are written back to disk, reads are done from memory). Inside this ramdisk I make a bunch of folders, /TMP, /TEMP, /TRASH. Saved the image back to disk and unloaded it. Now I have a 64kb image of a blank folder layout that I want.

Next I wrote a very small script that has imdisk load the 64kb file and then expand it out to 1GB as a RAMDISK (writes and reads are to memory, no saves are to disk). Mounts it as drive V:

I have changed my environmental variables to point TEMP and TMP to V:\TMP, V:\TEMP. Also changed the user TEMP folder to those locations. I'm also redirecting the temporary internet file location to those locations.


Now for the absolute crap that is the windows file caching. Windows file cache is sh!t, it's programmed to cache things from your user folder (documents, pictures, temporary internet files) over program data. In actual use this means windows file cache can't be trusted to perform properly for anyone not doing office stuff.
December 27, 2011 11:31:09 AM

palladin9479 said:
Now for the absolute crap that is the windows file caching. Windows file cache is sh!t, it's programmed to cache things from your user folder (documents, pictures, temporary internet files) over program data. In actual use this means windows file cache can't be trusted to perform properly for anyone not doing office stuff.


I agree wholeheartedly!

While you may have more technical knowledge than I do, it was not hard for me to see the advantages of a RamCache dedicated to my C drive Implemented by a well written program like SuperSpeed SuperCache insted of trusting windows to get the job done efficiently.

I am quite a fan of windows but I would never be so blind as to think it does everything perfectly!

It does seem that you have a pretty good implementation of a RAM disk. Is there any noticeable time is added to your restarts? While I'm sure if there is some for me, it isn't any thing noticeable. :sol: 
a c 415 G Storage
December 27, 2011 4:11:24 PM

palladin9479 said:
Windows file cache is sh!t, it's programmed to cache things from your user folder (documents, pictures, temporary internet files) over program data.
That's sure not how my Windows 7 system behaves. I don't even use the "Documents" folder tree to store the vast majority of my files, and they all get cached regardless of where they are at whatever folder depth on whichever drive. And every program I run is cached as well, as is very obvious from their startup times and by the fact that my continuously-running disk monitors show no disk activity whatsoever when re-running them after the first time they've been started.
December 27, 2011 11:07:54 PM

sminlal said:
That's sure not how my Windows 7 system behaves. I don't even use the "Documents" folder tree to store the vast majority of my files, and they all get cached regardless of where they are at whatever folder depth on whichever drive. And every program I run is cached as well, as is very obvious from their startup times and by the fact that my continuously-running disk monitors show no disk activity whatsoever when re-running them after the first time they've been started.



Your confusing superfetch with windows file caching subsystem, their two different components. The windows file cache is connected to the volume management subsystem, it only reads data that's been accessed recently and is weighted in favor of user's home folders. In practice it will always chose to retain data in your home folder over data elsewhere. Not only that but during disk reads the if determines that is a chance that the data in the cache is outdated, then it'll just reread the original data from the HDD anyway. Don't attempt to argue this, I've tested it pretty thoroughly.

Superfetch on the other hand is it's own service daemon that intercepts file system calls. It operates by reading all data into memory starting from most recently accessed and working it's way backwards. I haven't noticed any preference, but it doesn't distinguish temporary or scratch files from real data. It'll always try to cache your temporary internet files, and anything else it can get it's hands on.

Superfetch in Vista was complete crap, it's implementation wasn't synchronized with the windows memory subsystem and memory allocation calls would trigger the memory subsystem to page superfetch data out to to the page file. Superfetch in Windows 7 is much smarter, it knows to keep spare ram always available for memory allocation calls, sometimes it still ends up with superfetch cache being paged out but it's pretty rare now.

Windows file caching is sh!t, it's been sh!t for years. It's designed for office work, frequent access of small files that a user edits and works with. It's not very good at program data and absolutely atrocious at program binary's. A RAM Cache would be an incredibly useful tool, especially if you were able to manually tell it what data to maintain. You, the power user, know more then the OS ever will about what programs and data you will need to frequently access. All the OS can do is use last accessed dates and file types to try to guess what files you will use next.
a c 415 G Storage
December 28, 2011 2:33:25 AM

palladin9479 said:
Your confusing superfetch with windows file caching subsystem, their two different components.
That's funny, because when I read your post my first thought was that you had done the same thing.

I'm afraid we're just going to have to disagree. I do agree with you that a RAMDisk can be useful when you know exactly what you want to load into memory. And if you can do that initial load before you need the data and if the resulting disk activity doesn't slow down the other stuff going on then it can be a net gain. But in my experience Windows file caching works very well and relieves me of the need for that kind of micro-management.
December 28, 2011 3:12:09 AM

sminlal said:
That's funny, because when I read your post my first thought was that you had done the same thing.

I'm afraid we're just going to have to disagree. I do agree with you that a RAMDisk can be useful when you know exactly what you want to load into memory. And if you can do that initial load before you need the data and if the resulting disk activity doesn't slow down the other stuff going on then it can be a net gain. But in my experience Windows file caching works very well and relieves me of the need for that kind of micro-management.


It's known that windows file caching is designed around office automation tasks, its one of the key differences between the client and server versions of Windows. Simply put, there is no way for Windows to know ahead of time what needs to be in memory, the most they can do is to maintain a list of most commonly used files along with keeping the most recently used files in memory. Of course the caching system has issues with retaining that data for long periods of time. You can see this by opening 10MB file, writing a change to it and closing it. Wait 30 min, try to open it again, you'll notice that windows will read it from the disk again instead of from memory. The cache entry's time to live has expired and been flushed. Server version of Windows has a significantly longer time to live for cache entries, up to hours in some cases.

Really this has been gone over in detail already.
a c 415 G Storage
December 28, 2011 7:06:23 AM

palladin9479 said:
Of course the caching system has issues with retaining that data for long periods of time. You can see this by opening 10MB file, writing a change to it and closing it. Wait 30 min, try to open it again, you'll notice that windows will read it from the disk again instead of from memory. The cache entry's time to live has expired and been flushed. Server version of Windows has a significantly longer time to live for cache entries, up to hours in some cases.
OK, so I used my checksumming utility to checksum 4GB worth of files in an independent folder on one of my data drives. After an hour of other work I ran the utility again against the same 4GB worth of files. The drive didn't receive so much as a single I/O operation. Yes, it's Windows 7 (and not Windows Server 2008 or 2008 R2) and no, I've done nothing to tweak it to act like a server.

I'm still not buying your argument.
December 28, 2011 11:59:31 PM

sminlal said:
I'm still not buying your argument.


That's OK, because we aren't buying yours either. As far as I'm concerned you're just a stubborn old fool willing to condemn new technology without even giving it a fail trial. Very poor thinking, Very poor science. :pt1cable: 
a c 415 G Storage
December 29, 2011 4:25:31 AM

Idonno said:
Very poor thinking, Very poor science. :pt1cable: 
Hey, I tried the experiment. It confirmed what I said. If there's poor thinking or science going on, it's not at this end.
December 29, 2011 9:49:35 AM

sminlal said:
Hey, I tried the experiment. It confirmed what I said. If there's poor thinking or science going on, it's not at this end.

Really? because It was my impression that you hadn't tried any of the implementations of RamDisk or RamCache that "sminlal" or I mentioned. :kaola: 
a c 415 G Storage
December 29, 2011 2:37:22 PM

Do you think a RAMDisk is going to somehow reduce the number of disk I/O's below zero? Zero is as good as it gets. For the work that I do, where I don't have need to force specific files to reside into memory no matter what else happens, nor the inclination to spend the time fiddling with which set of files I load into it depending on what kind of work I'm doing, RAMDisk just isn't worth it.

I don't know why you guys have such a low opinion of Window's disk caching. I do a lot of program development so I run my system with a disk monitor always visible so that I can see "under the hood" to evaluate how well my code works. Over the last 2-1/2 years that I've been running Windows 7 I've become very familiar with how it behaves. And I just gotta tell you guys that your pronouncements of "horrible Windows caching" are just way off base.

Windows disk caching does NOT favour one set of files over another, unless the application specifically requests uncached I/O. It does NOT toss stuff out of the cache after a set amount of time. There is no specific cap on how much memory is used for caching. Where you guys picked up these ideas is beyond me.

EDIT - after thinking about this for a while, I'm wondering if you've come to your conclusions after using systems that don't have a lot of free memory, or from reading posts by other people in that situation. If you've got only a GB or so of free memory, then of course files aren't going to last in the cache nearly as long because I/O to new files will flush out the cached copies of older files. And naturally if you've allocated most of your free RAM to a RAMDisk then you'll see those kinds of deleterious effects on Windows' disk cache. Perhaps this is where the idea of a "time limit" for cached files came from.

But this thread started out talking about large memory systems, which I take to have several GB or more of free memory. If you don't have experience running your system that way, then perhaps your skepticism is understandable. But I gotta tell you that with plenty of free memory it works very well.
December 29, 2011 7:45:08 PM

@sminlal, Are you a member of the "The Flat Earth Society"? Yes they really do exist and they have lots of reasons and facts to back up their claim that the world is indeed flat. It may be a joke but, to someone who didn't actually know any better, some of their arguments might appear to be compelling.

Columbus was also given many compelling reasons to believe that the world was flat and I think that before the voyage even Columbus might have had at least a small amount of doubt in the back of his mind. Certainly some of his crew did! Just as certainly after the voyage both Columbus and his crew knew beyond a shadow of a doubt the "flat earth theory" was dead wrong!

While many people were willing to accept the fact that the earth was indeed not flat, some continued to hold on to the belief that it was. the biggest difference between these groups (Columbus and his crew VS flat earth believers) is one group knew better because they had tried it (the voyage) and the other group had not.

And just like the "The Flat Earth Society" It really doesn't matter how many reasons you throw out there to support your claim.
Both you and "The Flat Earth Society" have some good arguments, unfortunately "The Flat Earth Society" is wrong and so are you!

The biggest difference between us is I have tried it (RamDisk and RamCache) and You refuse to try it or even listen to people that have!

So yes, Very poor thinking, Very poor science! :pt1cable: 
December 29, 2011 7:58:52 PM

Reply to sminlal's edit:

sminlal said:
after thinking about this for a while, I'm wondering if you've come to your conclusions after using systems that don't have a lot of free memory


Quote:
Idonno Said:

I have 24GBs of ram 4 for my RAM pagefile, 8 for my RAM Cache and 12 left over for the OS. If you have enough ram try it. It definitely makes a bigger difference than I thought possible.

That is from my first post in this thread. :sleep: 

Look, I'm not saying that you don't have reason to believe what you believe. From reading your posts it's apparent that you have plenty of knowledge, more knowledge on this issue than I.

I AM saying that I came to these conclusions though actually trying these things. I really didn't expect SuperSpeed SuperCache (my RamCache) to make much difference either but, I can't deny what I have seen and felt. I don't even claim to completely know how the application SuperSpeed SuperCache works, only that it does work and it works extremely well!

Why, I really don't know but, I would have to guess it's because the program is extremely well written.

why don't you try it and stop all this silly arguing then maybe you can tell me! :sweat: 

Just because results (both "real world" & benchmark) fly in the face of everything you thought to be true, it doesn't mean the result's are wrong. It just means you have to adjust your way of thinking and maybe there is still more to learn. :sol: 
a c 415 G Storage
December 30, 2011 4:04:04 AM

Idonno said:
The biggest difference between us is I have tried it (RamDisk and RamCache) and You refuse to try it or even listen to people that have!
(sigh) OK, OK. I tried it. I downloaded the free 4GB version of the DataRAM product and tried it using three different workloads - a video alignment program I wrote, indexing and thumbnailing a directory of photos using Thumbs Plus, and checksumming the same directory.

The results: accessing the files on the RAMDisk was no faster than accessing the disk-resident files once they'd been cached by Windows. And having the RAMDisk software auto-save it's contents on shutdown resulted in really awful shutdown times. Startup times were longer too, but not by nearly as much - the software seems to be smart enough to load stuff in the background without delaying your ability to log in and do work - kudos.

What surprised me a bit (and I know you're not going to want to hear this), is that my tests actually ran just a tiny bit faster against files on disk cached by Windows vs. files loaded into the RAM drive. At first I thought this was just measurement error, so I created a side-by-side video of two runs just to be sure. You can see it here: http://www.youtube.com/watch?v=0bgIyKWlLbs I suspect that the difference may be due to the fact that Windows caching occurs at the file system level and therefore doesn't have the overhead of having to dive into the disk driver protocol stack.

Now it may well be that you've been able to optimize your workload better with a RAMDrive than Windows file caching can do. As I've said in my previous posts, if you know ahead of time what files you want fast access to, want to force them to always be resident in memory, and if the time to get them loaded into memory initially isn't a problem, then a RAMDrive can be a good solution. I have never tried to argue that it can't be under those kind of circumstances.

Compared to a RAMDrive, Windows file caching has two major disadvantages:

1) There's no guarantee that the files you want will stay cached if you're doing a lot of I/O and/or programs consume a lot of your free memory.

2) The initial access to files will be slower because they have to be read off the drive - so caching is of primary benefit for files that are accessed repeatedly.

That's not because Windows file caching is "sh!t", it's just the way caching works, whether it's disk caches, CPU caches, network caches, etc. It's a consequence of the caching strategy.

On the other hand, a disk caching strategy has some nice advantages too:

1) It doesn't reserve memory for use by caching only, which means you can run large, memory-intensive programs like video editors and have them take advantage of all your memory if you want them to.

2) It doesn't spend time loading files into RAM ahead of time (or saving them at shutdown time) even though you may not need to access those particular files during this session

3) It will cache any file, not just the ones you've decided to load manually. This works very well for people such as myself who routinely do a variety of tasks using different sets of files.

More than anything else, I want to make sure that people don't come away from this thread with the impression that "Windows file cache is sh!t". It's not. It does it's job very well, and none of the claims you guys have been making against it are true. I don't really have a problem with you saying that RAMDisk works well for you, what I take exception to is your attempts to convince people that anything else is worthless.

And with that, I'm outta here. I've spent enough time trying to convince you guys that Windows isn't the mess you think it is, and I don't really expect this post to change your mind. I just hope that other folks without an emotional investment can look at the exchange and come away with a balanced view.
December 30, 2011 6:45:56 AM

sminlal said:
(sigh) OK, OK. I tried it.

(sigh) NO, NO. YOU DIDN'T!

Again, I don't use ramdisk for caching. Like I said (for the third time now) " I have 24GBs of ram 4 for my RAM pagefile, 8 for my RAM Cache and 12 left over for the OS." and to be super clear because I've mentioned this 3 or 4 times now I use "SuperSpeed SuperCache" for a RamCache NOT just some generic RamDisk.

Why, because (also for the third time now) "SuperSpeed SuperCache" is a really well written program. It is designed specifically for the task of a dedicated RamCache. Read this and learn. It may also be important to note that the default size was 10GB's (at least for my system) and it still performs well at 8GB, however I don't know how well it would preform at 4GB which is only 2/5ths of my default setting.

I'm not saying 4GB's won't work well, just that I can only attest to 8GB. The default may differ depending on how much system ram you have.

You seem quite willing to do just about anything but use the program that I said works great for me even though you could have done so just as easily for FREE! Why is that?

Checking someones claims about how well one product works by testing a different product is pure Voodoo science. What's that even about?

Using a RamDisk to prove me wrong about "SuperSpeed SuperCache" would be equivalent to using "windows paint" to prove me wrong about "Adobe CS5" and quite frankly it stinks of a cheap shot by someone that knows they are wrong, but is to embarrassed to admit fault.

So again, Very poor thinking, Very poor NO science! WHAT-SO-EVER! :pt1cable: 

P.S. I don't understand what your video is trying to prove. since after the files are already cached in Windows' RAM cache they are in ram just the same as if they were in a RAMDisk. In ram is in ram. Whats your point?
December 30, 2011 10:39:31 AM

All the arguments seems to fade when you reserve 4 GB of your ram to put a page file on. That is lack of understanding, because it is just a waste of it.
December 30, 2011 3:02:26 PM

mikrev007 said:
All the arguments seems to fade when you reserve 4 GB of your ram to put a page file on. That is lack of understanding, because it is just a waste of it.


You are correct to a degree, except that as I stated in my first post "that completely eliminated all micro-stutter from my HD6870 crossFire set up." this was discussed further in my second post as well. Here also. That is the original reason I implemented a 4GB RamDisk for my paging file but, I have found there are other benefits as well. Whether I could achieve those "other" benefits with a smaller RamDisk-pagefile I don't yet know but, I suspect for most things other than my micro-stutter issue a 350MB Paging file would be sufficient.

The program I use "Dataram RAMDisk v3.5" is free forever for up to 4GB as well as being completely certified by windows for Win 7 64bit, so 4GB seemed to be as good a place as any to start.

I haven't had the app for very long. I have plenty of ram for most things I do. I have done some testing but since there is no trial period and all video micro-stutter has been eliminated, further testing has kind of taken a back burner to more important things.

So the lack of understanding may not entirely be on my part and that's OK I never claimed to be perfect either :sol: 
January 2, 2012 1:51:35 AM

I have had a little time to think about my last post and quite frankly I lowered my pride a little for the sake of peace on this thread. While peace in and of itself is a good thing, If it hinders someone from finding answers to questions or problems they came here seeking, IT'S JUST NOT WORTH IT!

The truth is I have a TOTAL understanding of the theory behind how a Widows pagefile is supposed to work which is exactly why after reading posts from people claiming success with the implementation of a RamDisk for their pagefile to fix the same issues that I had, it still took me a while to try it myself.

I don't claim to know why it worked only that beyond a shadow of a doubt it did! So for "mikrev007" to say that Implementing a RamDisk for a paging file is a waste of ram or a lack of understanding is completely wrong since people have had proven success solving certain issues and it shows a lack of understanding on his part alone.

What constitutes a waste of ram is completely subjective and depends to much on the users configuration, application and preferences among others. So a blanket "waste of ram" statement shows a complete lack of understanding on the part of "mikrev007" and his failure to further investigate or question the reasons and outcome.

As I stated before: "I am quite a fan of windows but I would never be so blind as to think it does everything perfectly!"
I have yet to see a perfect operating system. If Windows was perfect there would be no need for fixes and updates. The fact is Windows is constantly evolving to be better, safer and more stable.

This is going to be a fairly long post but, if you bear with me this might provide some useful information for some.

Background: A little I year ago I decided to build a fairly extreme PC. (specs can be found in "Member configuration")
My goal was to build a really good all-around PC for anything I chose to do, Including but not limited to A/V editing, Overclocking and Gaming. I'm not really that big into gaming or overclocking but when I do choose to do it I don't want to be too limited by my PC.

After completing my build and becoming comfortable with over-clocking this PC. I set out to find the fastest overclock I could run 24/7 safely and verified this with several burn-in apps. I then moved on to benchmarking to see how my new build stacked up with other PC's. It did very, very well especially in benchmark Apps that tested the entire system. The most notable was PassMark performance test. With PassMark my system benchmark was 7774.49 placing me 3rd of over 350,000 PC's benchmark-ed worldwide if you don't count the top 5 (EVGA Classified SR-2's with dual CPU motherboards) but, still 8th if you did.

I have since reset every thing to default speeds except for a mild CPU OC of 3.8MHz since I want this PC to last a long time but, I still occasionally play with the OC a little.

Issues: In spite of the fact that benchmarks aren't everything, this is a formidable PC in real world use. Certainly the most formidable all around (single CPU, Non-server) PC I have ever built/owned.

That said there were 2 issues that plagued this build since the beginning. The 1st was excessive micro-stutter from my HD6870 crossFire set up.

The 2nd was an occasional Audio/Video stutter when listening/watching different media files. the best way I can describe the stutter is that the audio sound is it's like a repetitive da,da,da,da..........approx 12-18 of these "da" sounds in about 2 seconds to with a quick video freeze for each "da" .

The first issue (micro-stutter) is not uncommon and in fact is likely more common in a dual HD6870 crossFire set up than any other video hardware set up I know of. Exactly how bad the issue is perceived to be is again very subjective. For me with 3 monitors in eyefinity with a resolution of 5670 X 1080 and a strong desire to use the highest possible game settings, I perceived the problem as extreme. Even though I don't game much this was supposed to be MY ultimate PC.

The second issue (occasional Audio/Video stutter) was very different from micro-stutter and only happened rarely (about once per day on avg.) and it could happen on playing any type of Audio or Video media on any player, web based or not it didn't matter.

Diagnosis: While it is commonly known why micro-stutter occurs in most instances, that was not the case with my 2nd issue (occasional Audio/Video stutter). Although I did find a few people that had described the exact issue, I could not find any fix for the problem. It is worth noting that the only commonality was Windows 7-64Bit.

I tried absolutely every thing possible to diagnose this problem with the exception of swapping my CPU, PSU and motherboard. Absolutely all other hardware was temporarily swapped out and checked one by one and wherever possible all voltages were checked with a multimeter, including the PSU and motherboard.

All drivers were checked. All codecs were checked and deleted/installed one by one wherever possible/necessary, the same goes for software. I did several complete formats and re-installs of my OS on several different hard drives in both raid and non-raid configurations.

All possible bios settings/options that I thought might even remotely affect this problem were tried as well!

Eventually I gave up on fixing this problem and was considering RMA'ing my motherboard. As annoying as I found this issue, it was fairly rare so I wasn't in a big hurry to be without this PC while waiting on a motherboard RMA that may or may not even fix the problem.

Solution: About six months ago while reading a post about a possible micro-stutter fix I came across my eventual solution: A RamDisk dedicated to pagefile as the only paging file on my system. I have found a few similar posts since then.

It is exactly because I DO NOT have a lack of understanding as how Windows is SUPPOSED to use the paging file and the causes of micro-stutter that I completely disregarded this fix for another five months and I might have never tried it had I not found a free win 7 64bit certified RamDisk App.

With nothing to lose I gave it a shot. To my surprise this completely eliminated all micro-stutter and the 2nd issue (occasional Audio/Video stutter) has been completely eliminated as well.

There is no other application that has made as big of an impact on the performance of my PC than RamDisk implemented in the way described. For most people (without those 2 specific issues) this type of implementation will not have as great of an effect.

Other Benefits: If you have enough ram you can eliminate your pagefile altogether, however it should be noted that every one at Microsoft whether they are a low level tech or a software engineer advises against this for a variety of reasons. Here is an excellent article on this issue: "Understanding the Windows Pagefile and Why You Shouldn’t Disable It"

Again If you have enough ram another option is to create a pagefile on a RamDisk while disabling the pagefile on all others. I would suggest a minimum of 350MB - 1GB. But only you can decide the optimal size for your configuration, usage and available ram. This will eliminate unnecessary writes to an expensive SSD and move them to even faster ram. While it is true that with lots of ram your expensive SSD's pagefile will not get written to much but, It will never not get written to at all, no matter how much ram you have, unless you locate it somewhere else preferably just as fast or faster like ram.

If you've found that you have more ram than you use anyway a RamDisk pagefile is an excellent way to put a small amount to good use and the idea that with lots of ram you wont need a paging file anyway is simply not true. A correct statement would be :"you won't need a paging file as much".

I am also testing a .75Gb RamDisk Pagefile in my old 32bit XP machine. Since only 3.25 of the total 4GB's is recognized anyway. Yet another good use. (so far, a little early to say for sure)

In Closing: I came to this thread to post about the tremendous benefits of using "SuperSpeed SuperCache" in response to the original posters question(s). The main reason I even mentioned using RamDisk as a Paging file when I posted my results was in the spirit of full disclosure since that was part of the configuration being tested in the sense that the pagefile that would normally be located on the tested drive was instead located on a RamDisk.

The controversy that created was somewhat understandable since I didn't go too in-depth into why I created a RamDisk-Paging file in the first place and what I did write seems to have been overlooked by some as well.

In my opinion there are other clear benefits to a RamDisk-Paging file as well. How great of a benefit that may be is clearly subjective as are most things in life and I urge any one that thinks "the generally accepted theory behind how a Widows pagefile is supposed to work" is absolute to read the link I posted.

Most importantly, I hope that some find value in what I wrote. If I had the opportunity to read a post like this when I first had the two separate stutter issues that would have saved me huge amount of trouble. :wahoo: 
January 2, 2012 2:38:37 AM

I can tell you that I use a RamDisk for the pagefile, and while I can't comment on any performance benefit I did it strictly to take the write thrashing off my system SSD. I also use the RamDisk for my temp/temporary folders and for my browser cache for this same reason. While write thrashing might not big as big of an issue as it was with older controllers I still see no reason whatsoever to make unneeded writes to my drives. Besides, it's a free program, and with 8GB of RAM most of it sits unused the majority of the time. I see no downside to it.

I do appreciate the conversation continuing. I will be giving Supercache a shot when I put my new rig together. I'll benchmark performance before and after and perhaps we can see how it'll affect 4x Vertex 3 60GB in RAID0.
January 2, 2012 2:53:13 AM

a4mula said:
and perhaps we can see how it'll affect 4x Vertex 3 60GB in RAID0.

See now your making me jealous! :lol: 

As far as "any performance benefit" with a RamDisk-pagefile, benchmarks didn't appear reflect any. I just think everything feels like it runs smoother even when I disregard the two more serious issues I had. This could just be my imagination but I don't think so. It feels to pronounced for that.
January 3, 2012 8:54:10 PM

So you argue that placing the page file on a ram disk is better than just turning off? It is not.

You lower the usable amount of memory. You then increase the commit limit by creating a page file in ram. That will lower your performance vs having no page file at all.

About the article you link to... You have misunderstod something. Moving the page file to a ram disk is essentially the same as disabling it (plus you add a overhead).
January 3, 2012 10:39:29 PM

mikrev007 said:
So you argue that placing the page file on a ram disk is better than just turning off? It is not.

You lower the usable amount of memory. You then increase the commit limit by creating a page file in ram. That will lower your performance vs having no page file at all.

About the article you link to... You have misunderstod something. Moving the page file to a ram disk is essentially the same as disabling it (plus you add a overhead).



Some programs won't work right without a page file though. Seems stupid but that's the NT memory subsystem for you.
January 4, 2012 12:14:58 AM

mikrev007 said:
So you argue that placing the page file on a ram disk is better than just turning off? It is not.

You lower the usable amount of memory. You then increase the commit limit by creating a page file in ram. That will lower your performance vs having no page file at all.

About the article you link to... You have misunderstod something. Moving the page file to a ram disk is essentially the same as disabling it (plus you add a overhead).


I believe I was very clear about all the reasons I use ramdisk in my last post.

If you believe it won't work out well for you don't use one. On the other hand it has worked out quite well for me with very noticeable positive results and no negative ones, So I will continue to use it.

All the negative reasons in the world haven't changed my positive results one bit! I guess my PC just isn't aware that my RamDisk/pagefile was supposed to have a negative impact.

So, you can stick to your negative reasons all you want, because to me, positive results are far more important. :bounce: 
January 4, 2012 2:24:53 AM

Idonno said:
I believe I was very clear about all the reasons I use ramdisk in my last post.

Nonetheless, you were very specific about that link and why you should not disable it. The reason you gave has nothing to do with the reasons given in that link.
January 4, 2012 2:29:29 AM

palladin9479 said:
Some programs won't work right without a page file though. Seems stupid but that's the NT memory subsystem for you.

There is nothing stupid about it. It is the few applications that are stupid. The developers are worried about running out of memory, so they check if there is a page file present. They cannot actually make use of it, they just feel more safe with it.

If you actually have such a game, see if that check cannot be disabled. It might just be the case.

Another thing... why reserve 4 GB for this?
!